In the morning—after compiling, packaging, and a steadying cup of coffee—she ran the signed driver package installation. Windows Defender asked for permission; User Account Control asked for grant; she watched the driver install events unfurl like a map. The Device Manager entry changed: the yellow triangle dissolved, replaced by a tidy icon and the words she craved: “Graphics Tablet — Pressure & Tilt Enabled.”
So she took a different route: WinUSB. The tablet enumerated as a WinUSB device; that meant that at least the OS could talk to it at a raw USB level. WinUSB was not glamorous—it exposed endpoints and transfers, bulk and interrupt pipe calls—but it was honest. It let user-mode applications send packets and receive replies without a kernel driver taking the wheel. She wrote a small, patient utility that opened the device by its VID and PID and queried its descriptors. The descriptor held a string she hadn’t expected: “ARTIST-0.9.” A firmware revision, perhaps. A hint. In the morning—after compiling, packaging, and a steadying
Mara opened the driver package again. This time, she read every line of the INF as if it were poetry, noting the service installations, the device class GUIDs, the registry values that set polling intervals and report descriptor sizes. She copied the manufacturer’s vendor certificate chain into a test machine she controlled, then created a local catalog (.cat) file that referenced the original signed binaries. It was delicate work—Windows checked catalog signatures against the driver files it referenced, but if the files were unchanged, the catalog would still validate. She avoided changing binaries, only extending the INF to include the missing PID and pointing the install directives to the same signed binaries. The tablet enumerated as a WinUSB device; that